Multiple Kernel SVM Based On Two-stage Learning

The method of multiple kernel learning is to solve the single kernel SVM for one or more times and the algorithmic complexity is very high for the big size of training samples.This implies that although multiple kernel learning method avoids the multifarious kernel and its parameters choice,multiple kernel learning method is usually very time-consuming and even difficult to implement when the scale of training samples is larger.Therefore,multiple kernel support vector machine based on two-stage learning algorithm is proposed in this paper,which can greatly reduce the time spend in model training.In order to show the proposed multiple kernel support vector machine based on two-stage learning algorithm for better,we first introduce the classical multiple kernel support vector machine algorithm.Then,we propose a new multiple kernel support vector machine based on two-stage learning algorithm and present the main theoretical analysis on the learning rate of multiple kernel support vector machine based on uniformly ergodic Markov chain samples.Finally,the performance of the proposed algorithm based on two-stage learning is tested on the public data set and compared with the other three multiple kernel learning algorithms.The numerical results show that compared to the three classical multiple kernel learning algorithms,the proposed multiple kernel support vector machine based on two-stage learning algorithm has better performance in three aspects of accuracy,total sampling and training time and the sparse of classifiers.